Search Images Maps Play YouTube News Gmail Drive More »
Sign in
Screen reader users: click this link for accessible mode. Accessible mode has the same essential features but works better with your reader.

Patents

  1. Advanced Patent Search
Publication numberUS20060274941 A1
Publication typeApplication
Application numberUS 10/603,215
Publication dateDec 7, 2006
Filing dateJun 26, 2003
Priority dateMar 28, 2003
Also published asUS7881561
Publication number10603215, 603215, US 2006/0274941 A1, US 2006/274941 A1, US 20060274941 A1, US 20060274941A1, US 2006274941 A1, US 2006274941A1, US-A1-20060274941, US-A1-2006274941, US2006/0274941A1, US2006/274941A1, US20060274941 A1, US20060274941A1, US2006274941 A1, US2006274941A1
InventorsKonstantin Zuev, Irina Filimonova, Sergey Zlobin
Original AssigneeKonstantin Zuev, Irina Filimonova, Sergey Zlobin
Export CitationBiBTeX, EndNote, RefMan
External Links: USPTO, USPTO Assignment, Espacenet
Method of pre-analysis of a machine-readable form image
US 20060274941 A1
Abstract
The present invention relates generally to an optical character recognition of machine-readable forms, and in particular to a verification of a direction of spatial orientation and a definition of a form type of the document electronic image. The goals of the invention are achieved by preliminarily assigning one or more form objects as elements composing a graphic image unambiguously defining its direction of spatial orientation. Similarly, one or more form objects are preliminarily assigned as elements composing a graphic image unambiguously defining its type. The direction of spatial orientation and the type of the form are verified via identification of said images. The models of graphic images either for verification the direction of spatial orientation or for defining the form type are stored in a special data storage means, one of the embodiment of which is form model description.
Images(2)
Previous page
Next page
Claims(19)
1. A method of machine-readable form pre-recognition analysis comprising
a filled in form image,
at least one form model description, containing spatial and parametric properties of at least one element of the said form,
processing at least the following steps:
preliminarily assigning at least one form object as an element of graphic image for identification of image direction of spatial orientation,
preliminarily creating at least one model of the said graphic image for identification of image direction of spatial orientation,
parsing image into regions,
form image spatial orientation direction verification, comprising at least the following steps:
detecting on the form image at least one element composing the graphic image for the image spatial orientation verification,
performing the said graphic image identification attempt to verify if the direction of the image agrees with that of the spatial orientation model,
performing graphic image turn from the current position to the preliminarily assigned direction on angle 90° and further returning to the previous step in the case of the image identification reliability level on the previous step being lower then the predetermined level thereof;
2. A method of machine-readable form pre-recognition analysis comprising
a filled in form image,
at least one form model description, containing spatial and parametric properties of at least one element of the said form,
processing at least the following steps:
preliminarily assigning at least one form object as an element of graphic image for identification of image form type,
preliminarily creating at least one model of the said graphic image for identification of image form type,
parsing image into regions,
form image type definition, comprising at least the following steps:
detecting on the form image at least one element composing the graphic image for form type definition,
performing a primary identification of the graphic image using the said model,
performing a profound analysis using a supplementary data in a case of multiple identification result of the said primary identification.
3. The method as recited in claim 1, wherein the direction of spatial orientation verification is performed via setting up and examining of hypotheses and the corresponding matching reliability estimation.
4. The method as recited in claim 2, wherein the form type definition is performed via setting up and examining of hypotheses and the corresponding matching reliability estimation.
5. The method as recited in claim 1, wherein the step of the form type identification is performed using minimum possible set of objects, defining the form type.
6. The method as recited in claims 1 or 2, wherein at least one object comprising the graphic image, is represented by non-text image.
7. The method as recited in claims 1 and 2, wherein at least one object comprising the graphic image, is represented by text image.
8. The method as recited in claim 7, wherein the text the said image is additionally recognized before the analysis.
9. The method as recited in claim 8, wherein the contents of the recognized text is used as an supplementary data in form type definition process.
10. The method as recited in claims 1 or 2, wherein the matching reliability estimation is performed on all steps of the pre-recognition analysis.
11. The method as recited in claims 1 and 2, wherein the at least one object comprising the graphic image, is represented by a group of form objects.
12. The method as recited in claim 6, wherein at least one element comprising the graphic image is the element of empty region type.
13. The method as recited in claim 6, wherein at least one graphic object comprising the graphic image is of dividing line type.
14. The method as recited in claim 2, wherein the profound analysis comprises at least
assigning on the form image at least one supplementary form element,
creating of profound analysis model using the said model of the said graphic image for preliminarily identification plus at least one said supplementary assigned form element,
performing a profound analysis of the form image using the said profound analysis model.
15. The method as recited in claim 2, wherein the profound analysis is performed using any other supplementary data.
16. The method as recited in claims 1 or 2, wherein the whole set of form objects is used to compose the graphic image for the direction of spatial orientation verification or for the form type definition.
17. The method as recited in claims 1 or 2, wherein the same graphic image is used for direction of spatial orientation verification and for the form type definition.
18. The method as recited in claims 1 or 2, wherein the said special model description is stored in the description of the form model.
19. The method as recited in claims 1 or 2, wherein at least one said form object, comprising the said graphic image is described in a form of alternative.
Description
REFERENCES CITED U.S. PATENT DOCUMENTS

6169822 Jan. 2, 2001 Jung 382/186
6148119 Nov. 14, 2000 Takaoka 382/203
6137905 Oct. 24, 2000 Takaoka 382/186
5592572 Jan. 7, 1997 Le 382/190
5471549 Nov. 28, 1995 Kurosu et al. 382/195
5235651 Aug. 10, 1993 Nafarieh 382/202
5031225 Jul. 09, 1991 Tochikawa et al. 382/203

REFERENCES CITED NON-PATENT DOCUMENTS

  • 1. “Proceedings of the 13-th International Conference on Pattern Recognition, Aug. 25-29, 1996, Vienna, Austria”. Vol. III, Track C, IEEE Computer Society Press, Los Alamitos, Calif., p. 681-685.
  • 2. 1. “Proceedings of the 13-th International Conference on Pattern Recognition, Aug. 25-29, 1996, Vienna, Austria”. Vol. III, Track C, IEEE Computer Society Press, Los Alamitos, Calif., p. 691-695.
  • 3. “Proceedings of the 13-th International Conference on Pattern Recognition, Aug. 25-29, 1996, Vienna, Austria”. Vol. III, Track C, IEEE Computer Society Press, Los Alamitos, Calif., p. 696-700.
  • 4. “Proceedings of the 13-th International Conference on Pattern Recognition, Aug. 25-29, 1996, Vienna, Austria”. Vol. III, Track C, IEEE Computer Society Press, Los Alamitos, Calif., p. 701-705.
  • 5. “Proceedings of the 13-th International Conference on Pattern Recognition, Aug. 25-29, 1996, Vienna, Austria”. Vol. III, Track C, IEEE Computer Society Press, Los Alamitos, Calif., p. 768-772.
  • 6. “Proceedings of the 13-th International Conference on Pattern Recognition, Aug. 25-29, 1996, Vienna, Austria”. Vol. III, Track C, IEEE Computer Society Press, Los Alamitos, Calif., p. 793-797.
BACKGROUND OF THE INVENTION

1. Field of the Invention

The present invention relates generally to an optical character pre-recognition of machine-readable forms, and in particular to bit-mapped image and one or more model matching methods, and also image spatial direction identification.

2. Prior Art

According to widely known methods of text pre-recognition a bit-mapped image is parsed into regions, containing text and/or non-text regions, with the further dividing said text regions into objects, containing strings, words, character groups, characters etc.

Some known methods uses preliminarily document type identification for narrowing a list of possible documents types, examined in an analysis of the document logical structure.

According to this group of methods the document type identification is an independent step of document analysis, forestalling logical structure identification. At that the document type and its properties list become defined up to the moment of defining the logical structure thereof. Or wise versa, a document structure identification may be an integral part of logical structure identification process. In this case the document type that fits closer the analyzed image is selected.

A spatial orientation direction verification is present in a number of documents.

In the U.S. Pat. No. 5,031,225 (Jul. 9, 1991, Tochikawa et al.) is disclosed a method of document image spatial orientation verification, using a preliminarily assigned character, to be found in the document. The found character is recognized to fit one of the 4 models thereof, corresponding with four possible directions.

The most reliably matching model indicates the orientation direction of the image.

The method causes a mistake in the case of possible different directions of text orientation to be present in the document. It also may cause mistake if the character is not reliably recognized after converting into image state.

In the U.S. Pat. No. 5,235,651 (Aug. 10, 1993, Nafarieh) the orientation direction of the image is estimated via setting up and accepting a hypothesis on the level of initial image units by analyzing the transition from dark points (pixels) and regions to light ones and wise versa. If the examined hypothesis is not accepted, the new one is set up, considering the image to be turned at 90° angle.

The method can't work if various orientation directions of text can be present on the form.

In the U.S. Pat. No. 5,471,549 (Nov. 28, 1995, Kurosu et al.) to define the image orientation direction the text characters are selected from the text one after another and are tried to recognize, supposing orientation direction to be 0°, 90°, 180°, 270°. The direction of the best matching is assumed as the right document image orientation.

The method can't work if various orientation directions of text can be present on the form as in the previous example.

In the U.S. Pat. No. 5,592,572 (Jan. 7, 1997, Le) the problem is solved by dividing the image into a large amount of objects, either of text or non-text types. Then the orientation of all initial objects is estimated via recognition of characters, with the further joining them into large ones and estimating the orientation thereof. Finally there is the only text object, covering the whole text field with the corresponding orientation estimation.

The main shortcoming of the method lies in that the orientation estimation is performed along with recognition of text portions, thus reducing the method output.

In the U.S. Pat. No. 6,137,905 (Oct. 24, 2000, Takaoka) and U.S. Pat. No. 6,148,119 (Nov. 14, 2000, Takaoka) the orientation direction is estimated by dividing the image into a plurality of regions, possessing various estimation weight coefficient. Then the orientation direction is estimated via the text recognition in the said regions. The total direction is estimated as a sum of particular ones together with their weight coefficients.

The shortcoming of the method is the low method output, depending greatly upon the recognition results.

In the U.S. Pat. No. 6,169,822 (Jan. 2, 2001, Jung) the predetermined portion of the text is parsed from the image and is performed (processed) recognition. In the case of recognition failure, the inference is made about the other orientation direction of the image.

To achieve the reliable result via the said method the large number of text portions are to be recognized. That surely reduces the method output.

SUMMARY OF THE INVENTION

One or more objects of the form are assigned thereon, composing graphic image, unambiguously defining its direction of spatial orientation. The said graphic image properties comprise a description of a special model for defining the direction of spatial orientation. Identification of the image with the said model the right direction of image spatial orientation is defined. The said model properties are stored in a special data storage means, one of the embodiment of which is the form image model description.

In the similar way one or more form objects are assigned thereon, composing graphic image, unambiguously defining its type. Additionally one or more form objects may be assigned, for the case of profound form type analysis, if two or more forms are close in appearance or in properties list. The graphic image properties comprise description of a special model for form type definition. The said model properties are stored in a special data storage means, one of the embodiment of which is a form model description.

After converting the form image is parsed into regions containing text images, data input fields, special reference points, lines and other objects.

The possible distortion, caused by the document conversion to electronic state, is eliminated from the image.

Objects, comprising the graphic image for spatial orientation verification, are identified on the form image. The orientation direction accuracy is verified and corrected if necessary.

The objects, comprising the graphic image for form type definition, are identified on the form image. The proper model is selected via identification of the said graphic image. In the case of multiple identification result, the profound analysis of the form type is performed. The profound analysis is performed in the similar way adding the supplementary objects to the graphic image and performing new identification.

The profound analysis is performed automatically or fully or partly manually.

BRIEF DESCRIPTION OF THE DRAWING

FIG. 1 shows a document with three assigned elements comprising graphic image.

FIG. 2 shows the graphic image, formed by assigned elements.

FIG. 3 shows some examples of graphic objects used as assigned elements comprising graphic image.

DETAILED DESCRIPTION OF THE INVENTION

The document logical structure examination requires dividing the document image into elements of different types. The single element of the document can contain its title, authors name, date of the document or the main text etc.

The composition of the document elements depends upon its type.

The document logical structure is performed by the following ways:

    • on the base of fixed elements location,
    • using a table or multi-column structure [1], [5], [6],
    • on the base of structural images identification [4],
    • via specialized methods for special documents types [3].

Methods of the first group requires fixed structural elements location and are used for fields mark out, i.e. image regions, containing elements for documents of standard form [2]. The exact elements location on the form may be distorted by scanning. The distortion may be of various kinds: shift, a small turn angle, a compression and stretching, a large turn angle.

All kinds of distortion are eliminated on the first stage of document image processing.

The coordinates of regions may be founded relatively to:

    • image edges,
    • special reference points,
    • remarkable form elements,
    • a correlation function, taking into account of all or a part of the listed above.

Sometimes, the distortion may be ignored due to its negligibility. Then the image coordinates are computed relatively to the image edges.

The most of the methods for form type identification uses special graphic objects that are reliably identified reference points, as black squares or rectangles, a short dividing lines composing cross or corner (FIG. 3) etc. Searching the reference points location combination as an image using the special models, the type of the analyzed form can be correctly defined.

The main technical result of the invention consists in gaining

    • universality of the pre-recognition analysis of machine-readable forms,
    • ability to process documents' images of more then one form type in one session,
    • ability to process documents images of different directions of spatial orientation,
    • ability to perform the pre-recognition process with high output.

The said technical result is achieved in the following way.

One or more objects (1) are assigned on the form, composing graphic image (2), unambiguously defining its direction of spatial orientation. The said graphic image properties are described in a special model used for defining the direction of spatial orientation. Identification of the said image via the said model the right direction of image spatial orientation is defined. The said special model properties are stored in a special data storage means, one of the embodiment of which is the form image model description.

In the similar way one or more form objects (1) are assigned, composing graphic image (2) on the form, unambiguously defining its type. Additionally one or more supplementary form objects may be assigned for profound form type analysis, if two or more forms are close in appearance or in properties list. The graphic image properties is described of an another special model used for form type definition. The said another special model properties are stored in a special data storage means, one of the embodiment of which is a form model description.

After converting to electronic state the form image is parsed into regions containing text objects images, data input fields, special reference points, lines and other objects.

The possible distortion, caused by the document conversion to electronic state, is eliminated from the image.

The objects, comprising the graphic image for spatial orientation verification, are identified on the form image. The orientation direction accuracy is verified and corrected if necessary.

The objects, comprising the graphic image for form type definition, are identified on the form image. The matching model is selected via identification of the said graphic image. In the case of multiple identification result, the profound analysis of the form type is performed. The profound analysis comprises creation of a new special model for form type identification containing primary special model plus one or more supplementary form objects. The image is performed a supplementary identification using new special model.

The profound analysis may be performed fully or partly automatically.

One or more form object, comprising the graphic image is described in a form of alternative.

REFERENCES CITED NON-PATENT DOCUMENTS

  • 1. “Proceedings of the 13-th International Conference on Pattern Recognition, Aug. 25-29, 1996, Vienna, Austria”. Vol. III, Track C, IEEE Computer Society Press, Los Alamitos, Calif., p. 681-685.
  • 2. 1. “Proceedings of the 13-th International Conference on Pattern Recognition, Aug. 25-29, 1996, Vienna, Austria”. Vol. III, Track C, IEEE Computer Society Press, Los Alamitos, Calif., p. 691-695.
  • 3. “Proceedings of the 13-th International Conference on Pattern Recognition, Aug. 25-29, 1996, Vienna, Austria”. Vol. III, Track C, IEEE Computer Society Press, Los Alamitos, Calif., p. 696-700.
  • 4. “Proceedings of the 13-th International Conference on Pattern Recognition, Aug. 25-29, 1996, Vienna, Austria”. Vol. III, Track C, IEEE Computer Society Press, Los Alamitos, Calif., p. 701-705.
  • 5. “Proceedings of the 13-th International Conference on Pattern Recognition, Aug. 25-29, 1996, Vienna, Austria”. Vol. III, Track C, IEEE Computer Society Press, Los Alamitos, Calif., p. 768-772.
  • 6. “Proceedings of the 13-th International Conference on Pattern Recognition, Aug. 25-29, 1996, Vienna, Austria”. Vol. III, Track C, IEEE Computer Society Press, Los Alamitos, Calif., p. 793-797.
Classifications
U.S. Classification382/181, 382/173
International ClassificationG06K9/34, G06K9/00
Cooperative ClassificationG06K9/00449
European ClassificationG06K9/00L1
Legal Events
DateCodeEventDescription
Feb 7, 2014FPAYFee payment
Year of fee payment: 4
Aug 26, 2013ASAssignment
Owner name: ABBYY DEVELOPMENT LLC, RUSSIAN FEDERATION
Free format text: ASSIGNMENT OF ASSIGNORS INTEREST;ASSIGNOR:ABBYY SOFTWARE LTD.;REEL/FRAME:031085/0834
Effective date: 20130823
Jun 13, 2004ASAssignment
Owner name: ABBYY SOFTWARE LTD., CYPRUS
Free format text: ASSIGNMENT OF ASSIGNORS INTEREST;ASSIGNORS:ZUEV, KONSTANTIN;FILIMONOVA, IRINA;ZLOBIN, SERGEY;REEL/FRAME:014727/0542
Effective date: 20030528